Experimental runs A repeatable synthetic scenario with declared inputs, seed, events and outputs.
A run is a bounded test of how an agent or organisation responds to a stated synthetic condition.
Sources and scenarios
Every run identifies the scenario definition, assumptions, model version and deterministic seed used to generate it. Sources may be authored hypotheses, approved reference material or previously promoted learning; provenance shows which was used. The service does not ingest live clinical, personal-account, manufacturing, emergency or control data.
Events and measures
Raw events are append-only evidence. Derived measures describe patterns such as detection time, response choice, adaptation and recovery. Repeating a seed makes a result reproducible; changing one declared condition supports comparison without hiding what changed.
Learning and promotion
A run can propose a learned delta, but it cannot silently rewrite the operating model. Reviewers examine the evidence and create a separate promotion record when a delta is accepted. The public dashboard exposes runs and provenance; protected research workspaces hold participant-only activity.